DYNAMICS AND ORIGIN OF THE AAD

The elemental systematics described above confirm that IMM lavas have consistently been derived by smaller degrees of melting than their PMM counterparts throughout the last ~30 m.y. More surprisingly, they indicate a marked decrease in overall degree of melting for both groups since ~14 Ma. Because there are no samples from 7- to 14-Ma seafloor, the timing and profile (whether it was gradual or incremental) of this decrease are unclear. However, the absence of discernible temporal or spatial gradients in either the 14- to 28-Ma Leg 187 data set or the 0- to 7-Ma near-axis data set suggests that the decrease may have been relatively abrupt and related to either or both of two significant tectonic events documented in tectonic reconstructions by Marks et al. (1999). Between ~14 and 12 Ma, both the deep central axis of the depth anomaly and (1–2 m.y. later) the IMM/PMM mantle boundary were relocated from Zone A to Zone B as a consequence of the northeastward migration of the 127°E Fracture Zone (e.g., Marks et al., 1999). At about the same time, the boundary between Segments B3 and B4 first developed as a permanent offset of the SEIR. This boundary is not a typical transform. Its position along axis is unstable (Marks et al., 1999), and it currently separates zones in which amagmatic, listric spreading has maintained opposite polarities for at least several million years (Okino et al., submitted [N2]).

We speculate that the new spatial coincidence of the AAD with cool mantle beneath the AADA led to decreased magma production and to the onset of pronounced and persistent amagmatic spreading including the separation of Segments B3 and B4. A second possible cause of the decrease in magma production can be related to an apparent broadening of the depth anomaly along axis. If one assumes that the depth anomaly became locked into the lithosphere close to the spreading axis (cf. Marks et al., 1999), then at ~8 Ma the depth anomaly began to expand to the west along the SEIR, almost doubling its along-axis extent by the present day (Marks et al., 1999; Gurnis et al., 1998). Expansion of the depth anomaly, reflecting cooling of the underlying mantle, would also account for decreased melt production. This hypothesis is consistent with dynamic models that indicate the relatively recent entrainment of residual slab material into the upwelling mantle beneath the AAD (Gurnis et al., 1998). If, however, the key assumption is incorrect and the depth anomaly did not lock in until the lithosphere was several million years old, then this hypothesis is difficult to sustain, as the apparent expansion of the depth anomaly is at least partly a dynamic phenomenon related to the spreading process.

Geophysical Evidence for Entrainment of a Residual Arc Source

Gurnis et al. (1998) and Gurnis and Müller (2003) proposed a detailed model for the origin of the AAD and the AADA. Based on dynamic plate motion reconstructions in a fixed mantle reference frame, they showed that the cold north-south elongate "source" of the depth anomaly coincides with the location of a long-lived, pre-100-Ma western Pacific subduction zone. The low mantle temperature beneath the AADA is inferred to derive from refractory subducted material that accumulated at the 660-km mantle discontinuity beneath this subduction zone. Furthermore, their mantle flow models incorporating this hypothesis indicate relatively recent (<20 Ma) entrainment of some of this refractory material along with overlying mantle wedge material into the upwelling upper mantle beneath the AAD. Cooling associated with this entrainment could potentially explain aspects of the evolution of the AAD, including the onset of its crenellated geometry.

One specific prediction of the entrainment model is a relatively recent reduction in mantle temperature beneath the AAD. The principal evidence that points to recent cooling of the mantle is discussed in the preceding sections. It includes the recent (<8 Ma) expansion of the depth anomaly and possibly the relatively recent (specific age unknown, but <14 Ma) increase in Na₂O contents of IMM lavas. However, the onset of both effects is rather more recent than the ~20-Ma initiation of the present crenellated plate boundary geometry that defines the onset of the cooling effect proposed by Gurnis and Müller (2003).

Despite these qualitatively consistent effects, the hypothetical entrainment of deep refractory mantle is questionable. Despite the compositional and thermal contrasts inherent in the tectonics of the AAD region, the Gurnis et al. (1998) and Gurnis and Müller (2003) models employ a uniform viscosity through the entire upper mantle. This simplified assumption has profound effects on model mantle flow patterns, and the results differ significantly from those of mantle flow models that explicitly incorporate variable mantle viscosity (West, 1997; West et al., 1997; Lin et al., 2002). In the three-dimensional models of West et al., incorporation of a cold, viscous region into the lower part of the upper mantle leads to preferential lateral inflow in the shallow mantle beneath the SEIR for a wide range of model parameters. This lateral flow replaces material removed by spreading while the cold region remains immobile due to its high viscosity. Lin et al. (2002) used dynamic variable-viscosity models to explicitly explore conditions under which cold mantle could upwell beneath the AAD. They concluded that for realistic viscosity structures, upwelling of cooler material is feasible only in relatively close proximity to thick continental lithosphere and is not sustainable in the long term. In this scenario, upwelling of cold residual arc material may have been initiated at ~45 Ma, when the present phase of relatively rapid spreading began, but entrainment of this material would have ceased by ~20 Ma. Thus, the degree of entrainment would diminish through time in the opposite sense to that inferred by Gurnis and Müller.

Geochemical Evidence for Entrainment

In principle, the entrainment model predicts that there should be an increasingly recognizable geochemical signal derived from the entrained subduction complex in progressively younger IMM lavas, but the nature of such a signal is difficult to predict. It could be dominated either by subducted material or by material from the original mantle wedge. If the latter, it could be characteristic of arc magmas derived from a "subduction-enriched" mantle wedge or it could have the "inverse" characteristics expected for a residual mantle wedge that was highly depleted as a consequence of arc melt production.

All IMM lavas have high Ba contents relative to PMM, and this distinction is more pronounced for near-axis lavas relative to those from Leg 187 sites (Fig. F8) (Christie, Pederson, Miller, et al., 2001). Taken at face value, these contrasts are in a sense predicted for a subduction-enriched signal, but Ba contents of IMM lavas are generally within the normal MORB range and the signature Nb depletion that typically accompanies Ba enrichment in volcanic arc settings is completely absent from the AAD region (C.J. Russo et al., pers. comm., 2004). PMM lavas are Ba depleted relative to their IMM counterparts and to normal MORB. Ba depletion could be consistent with inverse models invoking a depleted mantle wedge, but if that were the case, the strongest Ba depletion should be in IMM lavas and not, as observed, in PMM lavas.

Kempton et al. (2002) considered the potential geochemical consequences of subduction modification of a MORB mantle and proposed a multistage process involving initial depletion of a MORB mantle wedge, followed by subduction enrichment and radiogenic ingrowth. This model allows for a distinctive arc signature to be retained only in the relatively immobile Hf-Nd isotopic ratios of the wedge material with all other incompatible elements having been entirely removed in fluids. This allows for the relatively high Hf values of Zone A west lavas (13–15) to have been developed by radiogenic ingrowth of subduction-modified MORB mantle over periods of 500–1500 m.y. for reasonable parental Lu/Hf ratios. Based on this model, Kempton et al. (2002) argued that high-Hf Leg 187 IMM lavas from Zone A west are derived from such a subduction-modified mantle wedge and that this high Hf source has since expanded geographically, consistent with the entrainment model, to almost exclusively feed Zones B and C at present.

Unfortunately, two significant problems cast doubt on this interpretation. First, the entrainment model as proposed should impart a distinctive geochemical signature to the IMM/PMM boundary region that is not associated with the IMM province overall. This is not apparent from available data. Second, the model appears to require radiogenic ingrowth times that are significantly longer than allowed by the geodynamic constraints. These issues are discussed in the remainder of this section.

Confirmation of the entrainment model depends on a geochemical signature that is uniquely identified with the IMM/PMM boundary region and that expands along axis as entrainment proceeds. In terms solely of Hf, newly available data (Fig. F13) (Hanan et al., 2000a, 2000b, submitted [N1]) appear to confirm that relatively high Hf lavas (>13) occur throughout Segments B5 to B3, but their distribution is unconstrained to the west due to lack of data. In Segments B3 and B4, the diversity of Hf values reflects an overall extreme isotopic diversity (Pyle et al., 1992) that is not present in Zone A west. This diversity reflects, at least in part, a very low magma supply associated with the predominantly amagmatic spreading regime in these segments (Pyle et al., 1992; Christie et al., 1998) and is inconsistent with the Kempton et al. model in which the subduction-related signal is preserved solely in the Hf-Nd isotopic systematics.

On a broader scale, with the exception of several ultra-depleted lavas (Hf > 20), all AAD and Zone A lavas lie within the lower third of the known range for Indian Ocean MORB (Pearce et al., 1999; Kempton et al., 2002). This suggests that the locally high Hf values of the Leg 187 Zone A west sites and of Zone B axial lavas are a subset of a widespread IMM population and not a unique feature of the IMM/PMM boundary region. Hanan et al. (2000a, 2000b, submitted [N1]) have interpreted the higher Hf values among the broad array of IMM isotopic compositions in terms of mixing toward a widespread Archean subcontinental contaminant. In this respect it is noteworthy that a projection of the isotopic boundary trace would intersect the continent close to the eastern boundary of the Australian craton.

Finally, the time required by this model (>500 Ma) for radiogenic ingrowth appears to be too long to be consistent with the geodynamic constraints. According to the Gurnis et al. (1998) and Gurnis and Müller (2002) tectonic model, the original western Pacific subduction zone was overridden by the Australian plate and subduction ceased at ~100 Ma. Assuming a relatively old mean age for downgoing lithosphere of ~100 Ma and assuming that subduction operated at this location for ~100 m.y., the mean age of accumulated subducted material is unlikely to be significantly older than ~250 Ma, providing insufficient time for radiogenic ingrowth in a modified slab or mantle wedge with reasonable initial Lu/Hf.

The need for long radiogenic ingrowth times is particularly acute for several ultra-depleted, very high Hf IMM lavas from the AAD. Data of Salters (1996) and Hanan et al. (2000a, 2000b, submitted [N1]) include Hf values >20 at four sites in Segment B4 that are among the highest Hf values reported from the Indian Ocean. They are confined to regions of chaotic seafloor terrain formed by predominantly amagmatic, listric extensional faulting (Christie et al., 1998) where they are interspersed with low-Hf lavas (Fig. F13). They presumably represent a minor mantle component that is clearly manifested in lava compositions only when melting is limited to small intermittent batches. Although the known ultra-depleted lavas are spatially associated with the IMM/PMM boundary, Hanan et al. (2000a, 2000b, submitted [N1]) argued that an ultra-depleted end-member component is widespread as in normal Indian Ocean mantle. Regardless of its geographic extent, this component is unlikely to be derived from a relatively young upper Paleozoic–Mesozoic arc source, as there is insufficient time available for radiogenic ingrowth. Older potential source materials of both arc and cratonic origin are readily available farther to the west within or beneath the Australian continental crust.